The Politician and the Judge: Accountability in Government

By ERIC MASKIN AND JEAN TIROLE*

We build a simple model to capture the major virtues and drawbacks of making public officials accountable (i.e., subjecting them to reelection): On the one hand, accountability allows the public to screen and discipline their officials; on the other, it may induce those officials to pander to public opinion and put too little weight on minority welfare. We study when decision-making powers should be allocated to the public directly (direct democracy), to accountable officials (called “politicians”), or to nonaccountable officials (called “judges”). (JEL H1, H7, K4)

The premise behind democracy is that public any case, they have greater incentive than the decisions should reflect the will of the people. citizen to try to do so. After all, in any large But in most democracies, comparatively few society, a lone citizen will have strong tempta- decisions are made directly by the public.1 tion to free-ride, since her chance of actually More often, the power to decide is delegated to affecting policy is almost negligible.4 representatives. And there is a good reason for Another potential advantage of representative this delegation: representatives are usually ex- government is that it reduces the risk of “tyr- pected to do a better job.2 As specialists in anny by the majority.” Noting the dangers of public decision-making, they are more likely direct democracy,5 Madison (1787) writes: “It than the average citizen to have the experience, is of great importance ... to guard one part of the judgment, and information to decide wisely.3 In society against the injustice of the other part. Different interests necessarily exist in different classes of citizens. If a majority be united by a * Maskin: Institute for Advanced Study, Einstein Drive, Princeton, NJ 08540, and Princeton University (e-mail: [email protected]); Tirole: Institut d’Economie Industrielle, Manufacture des Tabacs, Bureau MF529-Bat. F, 21 allees de Brienne, 31000 Toulouse, France, and GREMAQ (UMR Sie«ye`s (1789) (see also the introduction to Bernard Manin, 5603 CNRS), Toulouse, CERAS (URA 2036 CNRS), Paris, 1997). and MIT (e-mail: [email protected]). We are grateful to the 4 Joseph Schumpeter (1942) puts it in characteristically participants of the CEPR-ESF conference on “The Institu- acerbic fashion: “The average citizen expends less disci- tions of Restraint” (Toulouse, June 24Ð28, 2000), the An- plined effort on mastering a political problem than he ex- nual Meeting of the European Public Choice Society (Paris, pends on a game of bridge.” April 18Ð22, 2001), the Southeastern International Trade 5 Summarizing the case against direct democracy, David and Economic Theory Conference (Miami, November 16Ð Butler and Austin Ranney (1994, pp. 17Ð18) write: 18, 2001), the Canadian Economic Theory Conference (To- ronto, May 24Ð26, 2002), the Modeling the Constitution The main arguments against holding referendums in Conference (Pasadena, May 16Ð17, 2003), and many sem- representative democracies include: (1) ordinary cit- inars for their observations on oral presentations, and to izens have neither the analytical skills nor the infor- Mathias Dewatripont, John Matsusaka, Stephen Morris, Ian mation to make wise decisions; (2) decisions by Shapiro, and two anonymous referees for helpful comments elected officials involve weighing the intensity of on previous drafts. preferences and melding the legitimate interests of 1 Ballot referendums constitute the largest class of deci- many groups into policies that will give all groups sions made through direct democracy, but even in the something of what they want; (3) decisions made by United States and Switzerland, where they are especially representatives are more likely to protect the rights popular, they touch on only a small fraction of public policy of minorities; and (4) by allowing elected officials to issues. be bypassed and by encouraging officials to evade 2 See, however, the sympathetic 14-page survey on di- divisive issues by passing them on to the voters, rect democracy in The Economist (December 21, 1996). referendums weaken the prestige and authority of Many have argued that once the digital divide is eliminated, representatives and representative government. e-voting will enhance the appeal of referendums. 3 The view that governments are better informed than We have already touched on points (1) and (3), and these citizens is emphasized in James Madison (1787) and Abbe« will figure prominently in our analysis. 1034 VOL. 94 NO. 4 MASKIN AND TIROLE: ACCOUNTABILITY IN GOVERNMENT 1035 common interest, the rights of the minority will “responsiveness,”8 it is in clear conflict with the be insecure.” rationale for representative democracy dis- But if representatives decide for the public, cussed above: that representatives can make what induces them to act in the public interest? better decisions than ordinary citizens.9 Further- In this paper we focus on two motivations in more, if minority rights are a concern, the abil- particular. ity to remove officials from office through First, we suppose that an official wishes to elections may give the majority too much power leave a legacy, i.e., she wants to be remembered to shape the government. for great things. Indeed, in our setup, it is not A constitution—a specification of who gets enough for the official that great things be done; to decide what—should strike a balance be- she wants to be the one who does them (al- tween these considerations. In this paper, we though we allow for the possibility that her compare constitutions from the standpoint of conception of what is “great” may differ from public welfare. We focus mainly on three dif- that of the average citizen, that is, she may be ferent modes for making decisions: (i) direct noncongruent with society).6 democracy, in which the public itself decides But the desire to use power to achieve certain through a referendum; (ii) representative de- ends is not the only motive we ascribe to the mocracy, in which an official subject to reelec- official. We also assume that she values being in tion (a “politician”) decides; and (iii) judicial office for its own sake, perhaps because she power, in which a nonaccountable official (a enjoys the perquisites that come with the job, “judge”) decides.10 We also show that, in our perhaps because she simply has a taste for wielding influence.7 The public can harness these two motives by 8 Opinions differed in the 1787 U.S. Constitutional Con- vention on the degree to which officials should be respon- making the official accountable, that is, by re- sive to public opinion. In this paper, however, we take the quiring her to run for reelection every so often. view of delegate George Clymer (see Charles Beard, 1913, Holding reelections creates two major potential p. 193), who wrote that “a representative of the people is benefits. First, it may induce an official who is appointed to think for and not with his constituents.” In otherwise inclined (because she is noncongru- doing so, we adopt the usual political-science definition of “representative democracy” as a system in which voters do ent with society) to act in the public interest. not instruct their representatives. The practice of instructing Because the electorate may not always be able representatives was widespread in the 18th century (e.g., to evaluate the official’s actions directly, this deputies to the Estates General in France, American states can be called the “moral-hazard-correcting” before the 1787 Constitution). The modern version of rep- resentative democracy gained acceptance in the 18th cen- benefit of accountability. Second, reelections tury in England, and with the 1787 Constitution in the may allow the electorate to “weed out” the United States and the 1789 revolution in France. See Manin noncongruent officials altogether. This can be (1997, Ch. 5) for more details. viewed as the “adverse-selection-correcting” 9 Received theories of democracy in political science, effect. e.g., Robert Dahl (1956), often stress the importance of repeated elections for making government responsive to the But accountability also carries with it two public. We argue that “moral-hazard-correcting” respon- serious possible drawbacks. In order to get re- siveness is beneficial, whereas simply carrying out what the elected, an official may choose an action, not public wants can be counterproductive. because it is right for society, but because it is 10 We use the term “judicial power” to refer to non- accountable officials because, in most democracies, popular. That is, she may pander to public opin- judges—at least, at the highest levels—are appointed rather ion. Although some might call such pandering than elected. In the United States, for example, all federal judges are appointed and with lifetime tenure. Most of what we say about “judges,” however, applies equally well to appointed government bureaucrats whose tenure is not ap- 6 For example, the “great things” she does might consist preciably affected by their actions. of favors for a particular interest group, in which case her Interestingly (and consistent with the theory developed legacy benefits could include what the interest group does here) elected judges in the United States appear to “pander” for her in return. more (and thus behave more like politicians) than their 7 The desire to hold office is a commonly assumed appointed counterparts. Specifically, Timothy Besley and motive in the political economy literature dating back to Abigail Payne (2003) analyze employment-discrimination Robert Barro (1973) and John Ferejohn (1986). It is our cases and find that in states where judges are subject to legacy motive that is nonstandard (although, we believe, reelection discrimination charges are filed at a higher rate. quite realistic). They attribute this finding to an incentive rather than a 1036 THE AMERICAN ECONOMIC REVIEW SEPTEMBER 2004 admittedly extremely simple basic model, the feedback” case we show, in Section III, subsec- welfare-maximizing constitution generally re- tion A, that the comparison of our three modes duces to one of these three modes (or a combi- of government turns on the strength of officials’ nation thereof). office-holding motive. When this motive is Of course, the requirement that officials run for strong, politicians always pander to public opin- reelection is not only the only form of account- ion11 (choose the popular action), and so the ability in political life. After all, most appointed best form of government (among the three) is officials are accountable to their supervisors. Elec- either direct democracy or judicial power (de- toral accountability, however, is more straightfor- pending on how much the electorate knows ex ward to analyze; it avoids the need to model the ante about the optimal action). When it is weak, motivations and beliefs of the supervisors. then politicians always act on their legacy mo- tivation, and elections offer some possibility to I. Overview screen out noncongruent officials. Hence, in that case, representative democracy dominates A. Outline judicial power (but still could be inferior to direct democracy if the public has a good idea In Section II, we set out a two-period model ex ante of the optimal action). These consider- with a homogeneous electorate. In each period, ations suggest that decisions of sufficiently there is a decision to be made between two great importance (ones for which the legacy possible actions. One action is “popular” in the motive is likely to dominate) are best taken by sense that the electorate believes it to be optimal politicians rather than by judges (although these with better than fair odds. The electorate will may also be decisions for which direct democ- either decide itself (direct democracy) or dele- racy fares better still). gate the decision to an official, who knows In Section III, subsection B, we expand the which action is optimal. Each official is either set of possible governmental systems to include congruent (i.e., she has the same preferences as any scheme in which the decision on whether to the electorate) or noncongruent with society, retain a first-period official depends on the first- although ex ante the electorate does not know period action. We argue that when the office- which case holds. She also places some weight holding motive is strong, no mechanism in this on holding office for its own sake. In the case of broader class dominates judicial power or direct delegation, the first-period official will either democracy (assuming that the public cannot stand for reelection just before period 2 (in the commit itself to a random election). When it is case of representative democracy), or remain in weak, however, a “hybrid” mechanism in which office automatically for the second period (the an unpopular official is replaced by direct de- case of judicial power). There may be some mocracy could be optimal. chance that, before period 2, the electorate Section III, subsection CÐSection III, subsec- learns whether or not the first-period decision tion F, consider several extensions. Section III, was optimal. subsection C, makes it costly for an official to In Section III, we analyze this model for the find out which action is best for her or for case in which there is no chance that the elec- society and argues that a politician is less likely torate learns about the optimality of the first- to incur that expense than a judge. We conclude period decision before period 2. For this “no that highly technical decisions are best taken by judges. Section III, subsection D, examines the issue of term lengths and points out that opti- mally they should balance the transition costs of selection effect, i.e., elected judges are likely to be more generous in awarding damages. (And among states with appointed judges, those where judges serve life terms—so need not worry about reappointment—have even fewer 11 As we have already noted, there is evidence from the discrimination charges filed.) Relatedly, Besley and Stephen empirical literature that elected and appointed officials be- Coate (2003) show that electricity prices paid by retail have differently. In addition to the articles cited in footnote consumers are lower in those U.S. states that elect regula- 10, work by Henning Bohn and Robert Inman (1996) and tors. The literature also provides evidence (reviewed in Andrew Hanssen (1999, 2000) suggests that elected judges Besley and Ann Case, 2003) that U.S. governors subject to (whom we would call “politicians”) are more independent term limits behave differently from those who are not. than appointed judges. VOL. 94 NO. 4 MASKIN AND TIROLE: ACCOUNTABILITY IN GOVERNMENT 1037 replacing officials (which make longer terms motive may induce elected officials to behave in desirable) and the electorate’s risk aversion congruence with the electorate’s wishes. Our (which argues for shorter terms). Section III, departures are the introduction of (i) pandering, subsection E, asks whether politicians or judges which is possible because the electorate has should have more discretionary power, and sug- only imperfect knowledge of the optimal policy, gests that politicians should have the edge be- and (ii) adverse selection which arises from the cause their decisions convey useful information possibility of officials’ noncongruence. to the public, unlike those of judges. Finally, The role of citizens’ initiatives as a form of Section III, subsection F, takes up the possibil- direct democracy is studied in Besley and Coate ity of campaign promises and argues that, de- (2000, 2003) and John Matsusaka (1992). The pending on how badly officials want to hold latter paper provides an empirical analysis of office, this can be either a boon or a bane. 871 Californian ballot propositions and shows In Section IV, we consider the case in which, that politicians tend to leave distributional is- with positive probability, the electorate learns sues to the public and to tackle “good govern- whether or not the first-period decision was ment” issues themselves. optimal before period 2. We show that, if the Specific points made in this paper relate to office-holding motive is strong, there are two disparate parts of the political economy litera- (mutually exclusive) alternatives to the full pan- ture. The notion that representative democracy dering equilibrium we obtain in Section III: gives rise to dynamic screening of politicians is either (i) officials always choose the optimal reminiscent of Juan Carrillo and Thomas Mari- action in the first period (if the probability of otti’s (2001) more general study of dynamic feedback is sufficiently high), or (ii) there is selection of leadership within a party. Papers by some chance that a noncongruent official will Persson et al. (1997), Mathias Dewatripont and pander (a good thing), but otherwise officials Tirole (1999), Jean-Jacques Laffont and David act on their legacy motive (if the probability of Martimort (1999), and Laffont (2000) present feedback is moderate). arguments for separation of powers that differ Finally, in Section V, we introduce the pos- from our pandering argument. sibility of conflict between majority and minor- Work on the rationale for unaccountabil- ity preferences and show that representative ity (as opposed to unaccountability’s conse- democracy may do a better job than either direct quences) is sparse. Much of the literature start- democracy or judicial power when the proba- ing with Kenneth Rogoff (1985) emphasizes the bility that minority preferences should prevail is benefit of having independent central bankers only moderate, but that judicial power is supe- with preferences different from those of the rior if the risk of minority oppression is suffi- electorate as a commitment device vis-a`-vis ciently large. We interpret the U.S. Constitution markets. Antione Faure-Grimaud and Denis as a device for switching control from elected to Gromb (2000) show that agency independence appointed officials when minority rights are stabilizes policies when bureaucracies are cap- jeopardized. tured by the industry. Matthew D. McCubbins et al. (1987) stress the ex ante control of B. Related Literature agencies. In its emphasis on welfare analysis, constitu- This paper borrows from several literatures. tional choice, and length of tenure, our paper is Our analysis of representative democracy builds perhaps most closely related to the literature on on the theory of elections as a disciplining de- posturing in corporate finance. This literature vice (Barro, 1973; Ferejohn, 1986; and chapters has argued that the threat of takeovers, liquida- 4 and 9 of Torsten Persson and Guido Tabellini, tion, or replacement induces managers to adopt 2000). These are models of political agency, in short-termist attitudes. They accordingly select which voters are ex post indifferent about whom they vote for,12 and in which the reelection

of adverse selection implies that voters learn nothing about an official from her behavior, and thus are indifferent about 12 Moral hazard is the only informational imperfection in whether or not to reelect her. Thus disciplining her relies on models following the Barro-Ferejohn tradition. The absence a particular resolution of this indifference. 1038 THE AMERICAN ECONOMIC REVIEW SEPTEMBER 2004 inefficient investments that pay off quickly ability as given and so does not broach the (e.g., Jeremy Stein, 1989; Dewatripont and tenure issue that figures so prominently in our Maskin, 1995; Ernst-Ludwig von Thadden, paper. 1995) or induce income and dividend smooth- ing (e.g., Drew Fudenberg and Tirole, 1995). II. The Basic Model The analysis of intraparty competition in Ber- nard Caillaud and Tirole (2002) also stresses the There are two periods, 1 and 2, and a pair {a, impact of entrenchment and competition for b} of possible actions in each period. (We give leadership in a mechanism-design framework. these actions the same labels in both periods Christopher Avery and Margaret Meyer (2000) only for notational convenience. In fact, they study the reliability of recommendations by an should be thought of as different between peri- evaluator who may be biased in favor of the ods, and not necessarily as describable ex ante.) agent to be evaluated and hired. An evaluator In this basic model, all voters have the same with career concerns may want to be tough to preference ranking14 of the two actions, but do preserve her reputation vis-a`-vis the principal. not know ex ante what this ranking is. The While career concerns always benefit the prin- optimal action—the action that voters would cipal in the basic model, Avery and Meyer favor if they knew their ranking—is drawn at identify conditions, in particular on the corre- random independently in each period. The prob- lation of the evaluator’s preferences over ability that it is a is p (Ͼ1⁄2), which is common time, under which reputational incentives are knowledge for the electorate. Thus, a is the harmful.13 “popular” action (in the sense that it would be A phenomenon somewhat akin to pandering chosen if the electorate did the choosing itself), arises in Stephen Morris (2001), who examines and b the “unpopular” one. The parameter p is a a model in which an informed advisor is sup- measure of how much the electorate knows posed to provide information to an uninformed about the issue.15 Two factors influence the principal with the same preferences. Morris value of p: technicality and familiarity. Techni- shows that, despite the coincident preferences, cality—the degree to which the issue requires the advisor may refrain from conveying her specialized knowledge, say of economics or sci- information truthfully if doing so might jeopar- ence—tends to reduce the value of p. By con- dize her reputation with the principal (e.g., in- trast, familiarity—how much exposure the crease the possibility that the principal thinks electorate has had to this sort of question—will that she is a “racist”). As in our model, this has raise p (we would classify issues that are largely the adverse welfare effect of possibly destroy- matters of basic values as “familiar”). ing useful information. We suppose that the electorate is interested in Finally, Brandice Canes-Wrone et al. (2001) maximizing its expected utility. For most of the analyze a career-concern model of political- paper (but see Section III, subsection D) we will incumbent policy choice in which, as in our assume that the electorate obtains a payoff of 2 model, the incumbent has policy expertise and if the actions chosen in both periods are opti- there is a positive probability of feedback before mal, a payoff of 1 if just one action is optimal, the reelection date. However, Canes-Wrone et and a zero payoff if neither is. That is, the al. focus on a quite different set of issues. In electorate is “risk neutral.” particular, the incumbent attempts to signal tal- Unless the constitution specifies direct de- ent (a high-quality official’s information is bet- mocracy, the period 1 decision is delegated to ter than that of a low-quality official) rather than congruence, and so the redistributive issues studied in Section V below do not arise. Also, the Canes-Wrone et al. model takes account- 14 We can accommodate heterogeneous preferences as long as they are such that maximizing overall welfare is the same as maximizing the welfare of the median voter. Thus, 13 Gilat Levy (2000) shows that a careerist judge over- in the basic model, we are supposing that the welfare of a turns precedent more than is socially efficient, as such minority cannot outweigh that of its complement. behavior is a signal of the judge’s ability. That paper also 15 In the case of a heterogenous electorate, p denotes the looks at the interaction between this incentive and endoge- median voter’s uncertainty about a;itisnot to be interpreted nous appeals. as the fraction of voters who think a is optimal. VOL. 94 NO. 4 MASKIN AND TIROLE: ACCOUNTABILITY IN GOVERNMENT 1039

16 an official. With probability ␲ (Ͼ1⁄2), the of- remaining in office in period 2 to that from choos- ficial is “congruent”; that is, her preference ing her preferred action in period 1. It could be ranking is the same as the voters would have, if affected by considerations beyond rent from of- they were fully informed. With probability 1 Ϫ fice-holding and legacy. For example, if a dis- ␲, the official is “noncongruent”;17 her pre- placed official derived some benefit from her ferred action differs from that of electorate.18 successor’s choosing her preferred action, then ␦ The parameter ␲ can be interpreted as both a would be somewhat smaller than the above for- description of the underlying pool of candi- mula suggest. In principle, ␦ could also be differ- dates, which for now we will take to be exog- ent for congruent and noncongruent officials enous—but see the discussion at the end of (although, for simplicity, we assume that it is not). Section IV—and as a measure of the elector- We suppose that before period 2 the elector- ate’s ability to screen officials. ate learns with probability q whether or not the We suppose that, unlike the electorate, an first-period action was the optimal one. With official knows which action is best for her (and probability 1 Ϫ q, the electorate learns nothing. which is best for society). This assumption re- Presumably the electorate will eventually find flects the official’s greater incentive to be well out about the optimality of this action, and so q informed (or, alternatively, her greater expertise can be thought of as a measure of the speed at and resources). which feedback accrues. To measure an official’s eagerness to hold Provided that direct democracy is not called onto power, we introduce a “notional discount for in period 2, the electorate can either main- factor” ␦. For example, suppose that the official tain the period 1 official in office (it is required obtains utility G from selecting her preferred to do so if the official was granted a two-period action,19 and utility R simply from being in term), or draw a new official from the pool of office (R may correspond to perks, prestige, candidates (i.e., elect a challenger). Because etc.). If the official’s rate of time preference is officials always choose their preferred action in reflected by discount factor ␤, then the trade-off period 2, the new official will select the optimal between second- and first-period payoffs is action with probability ␲.20

G ϩ R III. The No-Feedback Case (1) ␦ ϭ ␤ , G Let us assume in this section that the elector- since the official always selects her preferred ate obtains no feedback about the first-period action when in office at date 2. Notice that ␦ decision before period 2 (i.e., q ϭ 0). We first could be either greater or smaller than 1. compare the three benchmark institutions (di- More generally, the notional discount factor ␦ is rect democracy, representative democracy, and defined as the ratio of the official’s payoff from judicial power), next ask whether there exists an institution superior to the best of those three, and then go on to consider several extensions. 16 We will assume that constitutional dictates cannot be evaded. For a discussion of this assumption, see Barry R. A. Comparison of Institutions Weingast (1997). 17 With a heterogeneous electorate (see footnotes 14 and 15), the congruent official can be thought of as representing Direct Democracy (DD).—Under direct de- the interests of the median voter, whereas the “noncongru- mocracy, the decision is determined solely by ent” official represents some other group of voters. voters’ prior beliefs, and so the popular action is 18 We are supposing that the preferences of the congru- selected at each date. Thus, the electorate’s ex- ent official are perfectly aligned with the interests of the pected welfare is: electorate, while those of the noncongruent official are diametrically opposed. Without any change in the qualita- tive results, we could relax this assumption, so that the noncongruent official simply had a higher probability than 20 The model can be extended to an overlapping- the congruent official of being at variance with the elector- generations framework with two-period-lived officials. The ate in both periods. challenger’s incentives are then similar to those of the 19 The assumption that this utility accrues only if the incumbent one period earlier. Since our focus is on the official herself selects the action captures the legacy incumbent’s behavior, our simpler two-period model in- motivation. volves little loss of generality (but see footnotes 29 and 31). 1040 THE AMERICAN ECONOMIC REVIEW SEPTEMBER 2004

FIGURE 1. OPTIMAL CHOICE BETWEEN MODES

WDD ϭ 2p. voters learn nothing about the politician’s type from her first-period decision, since Judicial Power (JP).—Under judicial power, she always panders. Thus they are, in fact, the official need not worry about reelection and indifferent between reelecting and not re- so selects her preferred action in each period. electing her. Equilibrium welfare is The electorate’s expected welfare is therefore: WRD ϭ p ϩ ␲. WJP ϭ 2␲. Note that, generically, Representative Democracy (RD).—There are two cases: WRD Ͻ max ͕WDD, WJP͖.

(a) Strong office-holding motive (␦ Ͼ 1): When This is not surprising since, in this case, ␦ Ͼ 1, the politician values staying in office representative democracy is tantamount to through period 2 above selecting her pre- direct democracy in period 1 and to un- ferred action in period 1. In equilibrium, the accountable judicial power in period 2. politician selects the popular action a re- We conclude that, when ␦ Ͼ 1, RD is gardless of her true preferences and the dominated. The optimal choice between optimal action. Voters then reelect her if DD and JP turns on whether p Ͼ ␲ or and only if she chose a. We call this a full ␲ Ͼ p (see case ␦ Ͼ 1 of Figure 1). pandering (FP) equilibrium.21 Notice that

that, for ␳ Ͼ 0, the FP equilibrium of the text is the unique 21 FP is not the only equilibrium, but we claim that it is equilibrium in which the official does not randomize. the most reasonable one to focus on. The only other pure- Finally, although there are mixed-strategy equilibria in strategy equilibrium is that in which the official always which the official randomizes between a and b, each of chooses b (the unpopular action). However, this “unpopular these is non-Markovian in the sense that there are two pandering” equilibrium is not robust to a small perturbation different states (i.e., two realizations of the uncertainty to the pool of candidate officials. Specifically, suppose that about the official’s congruence and action a’s optimality) in we introduce a small proportion ␳ of officials who have which all actors have exactly the same preferences and yet weak office-holding motives (their discount factor is lower behave differently in the two states. These equilibria are, than 1): a fraction ␲ of these are congruent, the rest non- therefore, eliminated by the requirement that equilibrium congruent. In the Appendix, we show (see Proposition A1) strategies be Markovian (see Proposition A1). VOL. 94 NO. 4 MASKIN AND TIROLE: ACCOUNTABILITY IN GOVERNMENT 1041

(b) Weak office-holding motive (␦ Ͻ 1): When Observe that in the no-feedback case (q ϭ 0), ␦ Ͻ 1, the politician chooses her preferred the only benefit of accountability is the possi- action in period 1, i.e., there is no pander- bility of screening out noncongruent officials. ing. The electorate can now draw inferences That is, in this case, accountability does not about the politician from this choice. Spe- induce the official to act on behalf of the elec- cifically, the posterior probability that the torate (when ␦ Ͼ 1, the official panders; when politician’s preferences are congruent with ␦ Ͻ 1, she acts in her own interest). those of the electorate is We have implicitly been supposing that all decisions are equally important in the sense ␲p that they deliver the same potential legacy ͑Ͼ␲͒, ␲p ϩ ͑1 Ϫ ␲͒͑1 Ϫ p͒ payoff to the official. But we can readily accommodate variations in this payoff. Imag- ine that the legacy payoff is drawn each pe- when the official has chosen a in period 1; riod from a probability distribution with mean and G. Suppose that, in period 1 a politician faces an especially important issue— e.g., whether ␲͑1 Ϫ p͒ to go to war— generating potential legacy ͑Ͻ␲͒, Ј ЈϾ ␲͑1 Ϫ p͒ ϩ ͑1 Ϫ ␲͒ p payoff G , with G G. Then (if the second period issue is not yet known) the notional discount factor for this decision is when she has selected b. Thus, in equilib- rium, the official will be reelected if and G ϩ R ␦Ј ϭ ␤ only if she has chosen the popular action. ͩ Ј ͪ The electorate’s welfare is therefore G

WRD ϭ ␲͓1 ϩ p ϩ ͑1 Ϫ p͒␲͔ and therefore lower than that in (1) (we are focusing only on the first-period decision, be- ϩ (1Ϫ␲)p␲ϭ␲[1ϩ␲ϩ2p(1Ϫ␲)]. cause the magnitude of the legacy payoff does not affect second-period behavior). Such a de- Representative democracy now strictly cision would thus be less likely to generate dominates judicial power: The two generate pandering and therefore more apt to be best the same first-period behavior, but RD al- allocated to representative democracy rather lows the electorate to replace an official than to judicial power (note that the advantage whose behavior suggests noncongruence. of RD is not that it leads to better first-period Representative democracy dominates direct decisions than JP but that it provides an oppor- democracy if and only if tunity to screen officials). We conclude that, ceteris peribus, decisions generating large leg- ␲͑1 ϩ ␲͒ acy payoffs should be assigned to RD over JP (2) 2p Ͻ 1 Ϫ ␲ ϩ ␲2 (for such discriminatory assignment to be pos- sible, some aspect of the decision must be de- scribable in advance, e.g., that it entails whether ␦ Ͻ (see case 1 of Figure 1). Note that the or not to go to war). Of course, issues that ␲͑ ϩ ␲͒ 1 produce large potential legacy payoffs are likely “boundary” ϭ p between 2͑1 Ϫ ␲ ϩ ␲2͒ to generate large potential social payoffs as RD and DD in this case is below the well. Notice that if the social payoff from an corresponding boundary ␲ ϭ p between optimal first-period decision increases, the JP and DD in case ␦ Ͼ 1. That is, DD is boundary between RD and DD in Figure 1 (␦ Ͻ more likely to be optimal in case ␦ Ͼ 1 1) would move upwards, that is, RD would be than in case ␦ Ͻ 1. This is because in the less likely to generate greater expected welfare latter case, RD could be optimal even if than DD. This is because as the ratio of first- p Ͼ ␲ provided that the welfare gain from period to second-period payoffs rises, the com- having the opportunity to oust a possibly parative benefit from ousting a noncongruent noncongruent official is big enough. official after period 1 declines. 1042 THE AMERICAN ECONOMIC REVIEW SEPTEMBER 2004

B. Mechanism Design official has chosen the popular action) or else to hold a referendum on the second-period action Moving beyond the three systems DD, RP, (when the official has chosen the unpopular and JP, let us consider a more general system action). For ␦ Ͼ 1, improving on JP requires Ϫ for public decision-making in which an offi- reducing the difference xa xb to deter pander- ϭ ␦ ϭ cial makes the first-period choice and xa and ing. Probabilities xa 1/ , xb 0 constitute the xb are her probabilities of being reelected optimal choice. But contractibility is required after selecting the popular and unpopular ac- for such randomization. In its absence, DD or JP tions, respectively (our model is very simple, is optimal. and so xa and xb are the only instruments we have). To obtain an upper bound on social PROPOSITION 1: Expected welfares from di- welfare, we first assume that these probabili- rect democracy and judicial power are ties xa and xb are contractible (e.g., that it is somehow possible to write a constitutional WDD ϭ 2p and WJP ϭ 2␲. provision that if an official chooses a she will ϭ be reelected with probability xa). We then When q 0 (i.e., there is no feedback about the enquire whether the upper bound can be optimality of the official’s first-period deci- reached through an institution that does not sion), welfare from representative democracy require such contractibility. depends on the strength of the official’s office- Suppose that the official prefers action i over holding motive (i.e., whether ␦ Ͼ 1 or ␦ Ͻ 1). action j, where i, j ʦ {a, b}. She will select her When ␦ Ͼ 1, preferred action if WRD ϭ p ϩ ␲͑Ͻ max ͕WDD, WJP͖͒, Ͼ ␦͑ Ϫ ͒ 1 xj xi , and so representative democracy is dominated. and will pander if Judicial power is superior to direct democra- cies if and only if ␲ Ͼ p. When ␦ Ͻ 1, Ͻ ␦͑ Ϫ ͒ 1 xj xi . WRD ϭ ␲͓1 ϩ ␲ ϩ 2p͑1 Ϫ ␲͔͒, Note that, given her preferences, the official’s incentive to pander does not depend on and so representative democracy dominates whether she is congruent or not (this will no judicial power. Representative democracy longer be true when we consider the case q Ͼ is superior to direct democracy for values 0). That is, a congruent official who prefers a of ␲ and p such that (2) holds. has an objective function that is identical to When ␦ Ͼ 1, no other system (except those that of a noncongruent official who prefers a. requiring contractible probabilities) is better Following footnote 21, let us focus on than judicial power (if ␲ Ͼ p) or direct democ- Markov equilibrium, meaning that these two racy (if ␲ Ͻ p). When ␦ Ͻ 1, the hybrid scheme types of officials adopt the same equilibrium (RD Ϫ DD)—in which there is a first-period behavior. Clearly, randomization or pander- official who runs for reelection and, if defeated, ing by the official will not be welfare-maxi- is replaced by direct democracy—generates ex- mizing; in all the equilibria we discuss below pected welfare the official chooses her preferred action in period 1. Ϫ WRD DD ϭ ␲ ϩ 2␲p ϩ p2 Ϫ 2␲p2. If ␦ Ͻ 1 the only possible improvement on RD and DD occurs when ␲ Ͻ p. Then, for a range of parameter values, the welfare- If p Ͼ ␲, then the hybrid is better than maximizing system consists of RD in period 1 representative democracy and if p is not followed by an election in which the voters too much greater than ␲, it is the welfare- decide either to return the incumbent to office maximizing system (in particular, it also bet- (which, in equilibrium, will occur when the ter than direct democracy). VOL. 94 NO. 4 MASKIN AND TIROLE: ACCOUNTABILITY IN GOVERNMENT 1043

C. Costly Information Acquisition under judicial power, a conclusion that suggests that representative democracy will not accom- We have assumed that the official is perfectly modate technical decisions well. informed about the payoff consequences of ac- By contrast, under representative democracy, tions a and b. Suppose now that acquiring such we would expect politicians to expend consid- information in each period entails a private cost erable resources to ascertain the prior beliefs of c. That is, unless the official incurs c, she has no the electorate. Our model can be generalized to better information than the electorate. Assume allow the politician to be uncertain about the that the official’s gain G from selecting her electorate’s view. In this case, and if ␦ Ͼ 1, the preferred action satisfies politician will want to poll public opinion (pro- vided the cost of doing so is low enough) to (3) ͑1 Ϫ p͒G Ͼ c. know what the electorate actually favors. An unaccountable official would not incur such ex- This inequality is a necessary condition for penditures, which are socially wasteful in our an official under any system to acquire model. To sum up, accountability provides an information. incentive for wasteful information acquisition In the case of judicial power, this condition is and a disincentive for acquiring information also sufficient. However, it is not generally suf- about the optimal decision. However, because ficient under representative democracy. This is our model assumes that learning the beliefs of because information is less valuable to a politi- the public is unrelated to determining the opti- cian who risks losing her job if she acts on it. mal action, this summary perhaps exaggerates Put differently, pandering does not require the case against representative democracy. costly information (the implicit assumption here is that the cost of ascertaining public opinion is D. Term Lengths negligible relative to that of discovering the true consequences of a and b). We have been taking term lengths as exoge- To confirm this logic, note that, under RD, a nous. In particular, we have simply assumed congruent official will acquire information only that a judge serves for two periods. But we if might have supposed instead that the term is only one period long. (4) p͓G ϩ ␤͓G ϩ R Ϫ c͔͔ ϩ ͑1 Ϫ p͒G Ϫ c Actually, as the model stands, there is no difference (in welfare terms) between judges Ͼ pG ϩ ␤͓G ϩ R Ϫ c͔, serving one period or two: both deliver expected welfare 2␲. But one reason for this coincidence is that we have been assuming voters are essen- where the left-hand side is the official’s payoff tially risk neutral. Let us now suppose instead if she acquires information (and acts on it). The that they are risk averse. Specifically, let V(2) corresponding condition for a noncongruent of- (ϭ 2), V(1), and V(0) (ϭ 0) be the utilities ficial is always satisfied. It is easy to verify that corresponding to two, one, and zero optimal (4) is a more demanding constraint on G and c decisions over two periods, where V is a strictly than (3). That is, if (4) holds (3) does too. If, as concave function.22 Then, two-period terms still before, we take ␦ ϭ [␤(G ϩ R)/G], then (4) can produce expected welfare 2␲, but one-period be rewritten as terms now generate welfare 2␲2 ϩ 2␲(1 Ϫ ␲)V(1). Thus, because V(1) Ͼ 1 (from risk c 1 ϩ p␤ Ϫ ␤ aversion), one-period terms are better: it is less ␦ ϩ Ͻ ͩ Ϫ ͪ 1. risky to have two draws from the candidate pool G 1 p than just one. Of course, by this logic, we should replace Note that ␦ Ͻ 1 is no longer a sufficient condi- tion to avert pandering. 22 That V be concave is not the only reasonable possi- We conclude that information acquisition bility. If, for example, the payoff from an optimal decision about the consequences of public decisions is in the first period were enhanced by an optimal second- less likely under representative democracy than period decision, then V might be convex. 1044 THE AMERICAN ECONOMIC REVIEW SEPTEMBER 2004 officials as often as possible, which clearly does representative democracy is dominated),24 we not make sense. After all, there are setup costs have associated with every regime change, not the least of which is the learning-by-doing that a WRD ␴* ϭ Ͼ ␴* . new official must undertake. The optimal term RD 2 JP length will, therefore, strike a balance between 23 setup costs and risk aversion. Discretion confers a benefit under represen- tative democracy that is not available under E. Discretionary Power judicial power: it provides information that helps society weed out noncongruent officials. Judges and other nonaccountable officials Accordingly, elected officials should have more typically have narrower spheres of action and discretion than unelected ones. less discretionary power than accountable offi- cials, such as legislators. Thus, even such a F. Campaign Promises powerful institution as the U.S. Supreme Court can consider only the cases brought before it, In our model so far, the only information and, moreover, it is constrained (at least in under RD that voters have about an official at principle) to decide them according to the ex- the time of reelection is the action that she took isting law or the Constitution. By contrast, the in the first period. But in real elections, the U.S. Congress can set its own agenda and pass electorate typically has more to go on than just any law it wishes (although there may be a risk the candidates’ records. In particular, their cam- that the law will be struck down by the Court). paign promises may be informative. Let us One might ask whether such differences can be briefly examine how the analysis changes when explained in our framework. candidates can make promises during the elec- We cannot accommodate different degrees tion after period 1 about what they would do in of discretion in our basic two-action model, period 2. We suppose that, by that time, they but a simple extension allows us to do so. As know the optimal second-period action. before, let the social payoffs from the optimal If a promise implied no commitment, then, in and nonoptimal actions be 1 and 0, respec- our model, it would play no substantive role at tively. Introduce a status-quo alternative to all. Regardless of her true intentions, a candi- actions a and b. This status-quo action yields date would utter whatever helped her to get known social welfare ␴ ʦ [0, 1]. The question reelected, and so such pronouncements would is whether an official should be given the be worthless. Let us suppose instead that if a discretion to choose between a and b or be candidate promises to take an action, then, when required to stick with the status quo. The elected, she must carry out that pledge (perhaps answer depends on the value of ␴. Discretion because the loss of face from not doing so should be given to the official if ␴ is smaller would be too great). Assume furthermore that than some cutoff ␴*. challengers, as well as the incumbent, can make Under judicial power, the optimal value of such promises. ␴*is There are two cases depending on whether or not candidates are willing to carry out their WJP nonpreferred actions simply to get elected.25 If ␴* ϭ ␲ ϭ . JP 2 they are willing to do so, then, in effect, we obtain pandering in both periods (there cannot Under representative democracy and assuming that the politician does not pander (otherwise 24 We assume here that the status quo, if chosen, is selected for two periods. 25 In our basic model, an official always gets nonnega- tive utility (either R or G ϩ R) from holding office. How- 23 When officials are accountable an additional consid- ever, we can generalize this setup to allow for the possibility eration enters the picture: the length of term will affect their that an official who chooses her nonpreferred action derives incentive to pander. Because this factor is more compli- negative utility overall, in which case she would not prom- cated, we will not take it up here. ise that action simply to get elected. VOL. 94 NO. 4 MASKIN AND TIROLE: ACCOUNTABILITY IN GOVERNMENT 1045 be an equilibrium in which congruent and non- xa and xb denote the (sequentially rational) equi- congruent officials are separated because the librium probabilities that the official is retained latter officials would have the incentive to imi- in office when no feedback is obtained. In the tate the former merely to get elected). Appendix we show that if the fraction of can- In the case in which candidates pledge only didate officials with weak office-holding mo- their preferred actions, the electorate can infer tives (their ␦ is smaller than 1) is positive (even which action is optimal with high probability if if arbitrarily small), then, in the event of feed- there are sufficiently many candidates: the op- back, an official is reelected in equilibrium if timal action will be promised by approximately and only if she chose the optimal action in a fraction p of candidates; the other action by period 1 (see the proof of Proposition A2). approximately a fraction 1 Ϫ p.26 Thus, with There are four possible “types” of official in enough candidates, equilibrium will entail the the first period. In particular, let (C, a) corre- electorate selecting a candidate among those spond to a congruent official when a is the who have pledged the “more promised” action optimal first-period action (and hence the offi- (the action that a higher proportion of candi- cial’s preferred action). Let (N, b) refer to a dates have pledged to carry out). noncongruent official when the optimal action Thus, the welfare implications of campaign is b (so that the official prefers a). (C, b) and (N, pledges depend critically on which case we are a) are defined analogously. Let ⌬ denote the in. In the former case, pledges lead to the dete- difference between the official’s payoff from rioration of RD through pandering in each pe- choosing a and that from choosing b: riod. But in the latter, welfare in both periods ⌬͑ ͒ ϭ ϩ ␦͓ ϩ ͑ Ϫ ͒͑ Ϫ ͔͒ approaches the theoretical limit of 1 as the num- C, a 1 q 1 q xa xb ber of candidates grows (this gets at the intuitive ⌬͑ ͒ ϭ ϩ ␦͓Ϫ ϩ ͑ Ϫ ͒͑ Ϫ ͔͒ idea that the electorate benefits from political N, b 1 q 1 q xa xb competition). ⌬͑ ͒ ϭ Ϫ ϩ ␦͓ ϩ ͑ Ϫ ͒͑ Ϫ ͔͒ N, a 1 q 1 q xa xb IV. Feedback ⌬͑C, b͒ ϭ Ϫ1 ϩ ␦͓Ϫq ϩ ͑1 Ϫ q͒͑x Ϫ x ͔͒. A. Forward-Looking Pandering a b

Let us now assume that parameter q is posi- We have listed the four ⌬s in descending order for tive, so that with positive probability the elec- the case ␦q Ͻ 1. When ␦q Ͼ 1, then ⌬(N, a) Ն torate learns before period 2 whether the ⌬(N, b); the ranking is otherwise unchanged. It first-period action was optimal. Expected wel- will make the analysis more interesting to sup- fares under direct democracy and judicial power pose throughout that ␦ Ͼ 1.27 are unchanged: under direct democracy feed- There are three possible equilibria,28 two of back is irrelevant since optimal decisions are which are mutually exclusive. independent across periods; under judicial power, it is also irrelevant, since the official Full Pandering Equilibrium.—As in Section cannot be voted out. Thus, as before, III, full pandering entails that the official always selects the popular action in equilibrium and is WDD ϭ 2p reelected if and only if she adheres to equilib- rium. The necessary and sufficient condition for and such an equilibrium to exist is that ⌬ (C, b) Ն 0 ϭ ϭ when xa 1 and xb 0, that is, WJP ϭ 2␲.

Next consider representative democracy. Let 27 For ␦ Ͻ 1, the possibility of feedback does not alter the official’s first-period behavior: she always chooses her preferred action. But feedback allows the electorate to learn 26 If there is a cost to making campaign pledges, then with certainty whether or not the official is congruent. some candidates who prefer the nonoptimal action may 28 Here we are assuming that, as in footnote 21, a small refrain from making pledges at all, since they are unlikely to proportion of officials who have weak office-holding mo- be elected. tives are introduced and this proportion is then sent to zero. 1046 THE AMERICAN ECONOMIC REVIEW SEPTEMBER 2004

␦͑1 Ϫ 2q͒ Ն 1. are an FLP equilibrium in which, if there is no feedback, the electorate randomizes over reelec- Thus, the official must have a strong office- tion (this limit always exists for this range of holding motive, and feedback must be suffi- parameter values) and a full pandering equilib- ciently slow. As noted in the previous section, rium (if ␦(1 Ϫ 2q) Ն 1). FP makes representative democracy unattrac- tive, since welfare is then p ϩ ␲, which is less Partial Pandering.—When than max {WDD, WJP}. ␦q Ͻ 1, Forward-Looking Pandering.—In this next kind of equilibrium, the official selects the op- then FLP is no longer an equilibrium. The only timal action in the first period regardless of her equilibrium possibility (other than full pander- own preference. That is, types (C, a) and (N, a) ing) is that types (C, a) and (N, b) select action select action a and types (C, b) and (N, b) a, type (C, b) selects action b, and type (N, a) choose action b. In the absence of feedback, the selects her preferred action b with probability y official is always reelected. A necessary and and panders with probability 1 Ϫ y, where sufficient condition for such an equilibrium to exist is that ⌬(N, b) Յ ⌬(N, a), or as we noted ␲͑1 Ϫ p͒ ϭ ␲, above, ␲͑1 Ϫ p͒ ϩ ͑1 Ϫ ␲͒py ␦q Ն 1. or We call the official’s behavior in this equilib- rium forward-looking pandering (FLP) to re- 1 y ϭ Ϫ 1. flect the fact that she is pandering not to the p current electorate but rather to voters of the future who may have received feedback about her first-period performance.29 FLP illustrates Thus (N, a) sometimes panders, but the other the disciplinary role traditionally attributed to three types never do. elections in representative democracy (indeed, To compute social welfare in this partial it constitutes a pure moral-hazard-correcting ef- pandering (PP) equilibrium, note that, without fect; there is no correction of adverse selection). feedback, the electorate is indifferent between It generates welfare replacing the official and keeping her, regard- less of her first-period behavior (because, after 1 ϩ ␲ ͑ϾWJP͒. either choice of action, the conditional proba- bility that she is congruent is ␲). So, without In the Appendix (Proposition A2) we show that feedback, second-period welfare always equals if q␦ Ͼ (1 ϩ ␦)/2, the unique limit of perfect ␲, and overall welfare is therefore: Bayesian equilibria when a proportion ␳ of of- ficials with weak office-holding motives is in- ͓␲ ϩ ͑1 Ϫ ␲͒p͑1 Ϫ y͔͒ ϩ ␲ troduced and then sent to zero is the FLP equilibrium in which an official is reelected if ϭ͓␲ ϩ ͑1 Ϫ ␲͒͑2p Ϫ 1͔͒ ϩ ␲ Ͼ WJP. and only if either (i) there is feedback and she has chosen the optimal action or (ii) there is no feedback and she has chosen action a. Further- The PP equilibrium generates greater welfare more, Proposition A3 shows that if (1 ϩ ␦)/2 Ͼ than JP in the first period: either an official q␦ Ͼ 1, then the only two possible limits as ␳ 3 0 chooses her preferred action (as in JP) or, if noncongruent [more specifically, if of type (N, a)], possibly panders (which enhances welfare). 29 If we instead adopted an overlapping-generations PP also generates greater welfare than JP in the framework, the official would no longer always choose the optimal action in equilibrium, even in the case ␦q Ն 1. second period: if there is no feedback after the Instead we would obtain an equilibrium very much like first period, the two are the same. But if there is what we call partial pandering (see footnote 31). feedback, PP improves the likelihood of a con- VOL. 94 NO. 4 MASKIN AND TIROLE: ACCOUNTABILITY IN GOVERNMENT 1047 gruent official, since noncongruent officials are adverse selection. With moderately fast feed- more likely to be thrown out of office. Notice back (␦q Ͻ 1, q Ͼ 0) the partial pandering that the PP equilibrium incorporates both equilibrium exists and to some extent counter- adverse-selection-correcting and moral-hazard- acts both moral hazard and adverse selection. correcting effects. In the Appendix, we show (Proposition A4) that when ␦ Ͼ 1 and q␦ Ͻ 1, B. Performance Measurement Within one limit of perfect Bayesian equilibria as the Government proportion ␳ of officials with weak office-hold- ing motives goes to zero is a PP equilibrium.30 We have seen that an increase in the quality The only other possible limit [if ␦(1 Ϫ 2q) Ն 1] q of performance measurement can boost the is the full pandering equilibrium. case for representative democracy. Indeed, there To summarize, we have: is a potential advantage from raising q beyond the direct benefit of inducing a forward-looking PROPOSITION 2: Suppose that officials’ of- or partial pandering equilibrium. We have been fice-holding motives are strong (␦ Ͼ 1). Feed- taking the pool of potential officials as given back (q Ͼ 0) creates scope for socially exogenously. But in practice it will depend on beneficial forward-looking pandering31 (when the cost a candidate incurs from running for ␦q Ն 1) or partial pandering (when ␦q Ͻ 1) and office and her expected payoff from holding eliminates full pandering when ␦(1 Ϫ 2q) Ͻ 1. office. Notice that moving from full pandering These three sorts of equilibria are the only to forward-looking or partial pandering raises limiting possibilities when a proportion of offi- the payoff for congruent officials and reduces cials with weak office-holding motives is intro- that for noncongruent officials. Thus, through duced and then sent to zero. the self-selection it promotes, such a move is likely to improve the composition of the pool. Thus, the equilibria of this model exhibit the We have assumed that q is exogenous, but full range of potential benefits from account- institutional design in practice affects perfor- ability. In the absence of feedback (q ϭ 0, mance measurement. Of particular interest here Section III), the reelection process may help is the creation of independent evaluation boards counteract adverse selection but does nothing to (e.g., the U.S. General Accounting Office) to solve the moral hazard problem: either ␦ Ͼ 1, in write detailed opinions about the performance which case the politician panders and represen- of elected officials (e.g., the budget designers). tative democracy is dominated; or ␦ Ͻ 1, in Such evaluation boards do not share control which case behavior is the same as without with elected officials. Rather, they perform an accountability, and elections merely offer the advisory or monitoring role. They can be prospect of removing noncongruent officials. viewed as raising the quality q of the elector- With rapid enough feedback (␦q Ͼ 1) the ate’s performance measurement. forward-looking pandering equilibrium exists Although we often take such institutions for and completely solves the moral hazard prob- granted, we might enquire into their rationale: lem in the first period, but does not help against why is the monitor most often32 an unaccount- able body? And why are the evaluated officials generally accountable? 30 In Section III, we rule out equilibrium in which the official randomizes by appealing to Markov equilibrium The answer to the latter question is straight- (see footnote 21 and Proposition A1). Although partial forward in our model: improved performance pandering entails randomization by the official, it is consis- measurement pays off only if the evaluee is tent with the Markovian requirement because, once q Ͼ 0, accountable. Indeed, the effect of performance the different types of official no longer have the same measurement grows with the degree of preferences. 31 In footnote 29, we noted that if we used an overlap- accountability. ping-generations model instead of our two-period, once-off The former question is more challenging. A framework, we would no longer obtain a FLP even in the tentative answer could be that if an official has case ␦q Ͼ 1. Instead, we would get a PP equilibrium in which (i) types (C, a), (C, b), and (N, b) choose their preferred actions, (ii) type (N, a) panders with probability y, and (iii) the official is reelected only when there is feedback 32 To be sure, there also exist Congressional investiga- that she chose the optimal first-period action. tion committees in the United States. 1048 THE AMERICAN ECONOMIC REVIEW SEPTEMBER 2004

the incentive to pander to the electorate, then so TABLE 1—OFFICIALS’ PREFERRED ACTIONS might an accountable monitor. In that case, the introduction of monitors would do little to ex- Type of official pose panderers. Case MmW Majority should prevail aba V. Tyranny of the Majority Minority should prevail abb

We now relax the assumption of a homoge- neous electorate in order to investigate minority politics. We will argue that, relative to JP, rep- overall net loss L Ͼ 0 relative to action b. B can resentative democracy gives more weight to the be interpreted as the social loss when the mi- majority and is therefore more likely to under- nority blocks a socially desirable move, value the minority interests. whereas L is the social loss from a move that The tyranny of the majority is an old theme of oppresses the minority. political thought. We have already quoted Mad- Finally, we assume that there are three types ison on this point. In The Federalist Papers of officials: those who are congruent with the Alexander Hamilton proposed that the judiciary majority (labeled “M” and having probability ␲ should control the encroachments and oppres- M); those who are congruent with the minority ␲ sions of the representative body and that insu- (labeled “m” and having probability m); and lating it from accountability is important for this those who balance the two groups’ interests and role: so have preferences in line with social welfare ␲ (labeled “W” and having probability W). The preferences of the three types of officials are If, then, the courts of justice are to be considered as the bulwarks of a limited summarized in Table 1. Constitution against legislative encroach- Let us look at the welfare implications of our ments, this consideration will afford a three benchmark institutions: strong argument for the permanent tenure Direct democracy: Under DD, the majority of judicial offices, since nothing will con- always prevails, and so per-period social wel- tribute so much as this to that independent fare is spirit in the judges which must be essen- tial to the faithful performance of so ar- xB Ϫ ͑1 Ϫ x͒ L. duous a duty. Judicial power: A nonaccountable official se- Still, despite the risks that RD poses for minor- lects her preferred policy, and so per-period ity rights, we will suggest that it safeguards welfare under JP is them better than direct democracy (recall Butler ␲ ͓ Ϫ ͑ Ϫ ͒ ͔ ϩ ␲ ͓ ͔ ϩ ␲ ͓ ͔ and Ranney, 1994, on the pitfalls of democracy; M xB 1 x L m 0 W xB . see footnote 5). We introduce a variant of the basic model in As one would expect, direct democracy fares which there are two groups, the majority and the better relative to judicial power when the loss minority. To simplify the analysis, we assume from the externality is small (L small) or that the majority knows that action a is its op- unlikely (x large), and when officials tend to ␲ timal choice. Similarly, the minority knows that be biased toward interest groups ( W low). action b is best for it. What is not known is Representative democracy: The analysis of which action maximizes overall welfare. With RD is similar to that of Section III. Let us begin probability x the majority should prevail, be- with the pandering case ␦ Ͼ 1. In that case, even cause action a imposes only a small negative a W official who knows that the minority should externality on the minority; let B Ͼ 0bethe prevail prefers to cater to the majority in period overall social benefit of action a over action b in 1. As in Section III, RD delivers the same that case (we normalize social welfare under outcome as DD in period 1 and the same as JP action b to be 0). With probability 1 Ϫ x, the in period 2. It is therefore dominated. minority should prevail: action a imposes a Next suppose that ␦ Ͻ 1, so that there is no large externality on the minority and generates pandering. Action a is chosen by an M official VOL. 94 NO. 4 MASKIN AND TIROLE: ACCOUNTABILITY IN GOVERNMENT 1049 and, with probability x,byaW official. Action official, but RD does not overweight m officials’ b is selected by an m official and, with proba- influence as does JP. Finally, for high values of bility 1 Ϫ x,byaW official. Thus, the majority x or B/L, DD is optimal (since, in this section, will not reelect an official who has chosen b, we assume away imperfect knowledge about and so representative democracy weeds out m the optimal action on the part of the elector- and (less effectively) W officials. ate, we might as well let the majority decide directly if the minority stands to lose rela- PROPOSITION 3: For all ␦, (i) there exist tively little). thresholds x* and x** (x* Ͻ x** for ␦ Ͻ 1, x* ϭ x** for ␦ Ͼ 1, where 0 Ͻ x* Յ x** Ͻ 1) such The Constitution as a Decision-Allocating that if x Ͻ x*, JP is optimal, if x* Յ x Յ x**, Device.—Proposition 3 suggests a possible in- RD is optimal, if x** Ͻ x, DD is optimal, where terpretation of the U.S. Constitution. Let us x is the probability that the majority should suppose that x is rarely so high that DD is prevail. (ii) Similarly, there exist thresholds desirable. Then, it becomes desirable to distin- B * B ** B * B ** guish between those cases in which x is (mod- ͩ ͪ and ͩ ͪ ͩͩ ͪ ϭ ͩ ͪ if ␦ Ͼ 1ͪ such L L L L erately) high (so that RD is optimal) and those B B * B * in which x is low (JP is optimal). that if Ͻ ͩ ͪ , JP is optimal, if ͩ ͪ Յ The Constitution provides a means of making L L L this distinction operational. If a decision bears B B ** B ** B Յ ͩ ͪ , RD is optimal, if ͩ ͪ Ͻ , DD on some Constitutional guarantee, this is a sign L L L L that x is low, i.e., that a minority’s rights are is optimal, where B/L is the ratio of the welfare in jeopardy. And, indeed, the decision mech- loss from minority blocking to that from major- anism in such a case is to assign the decision ity oppression. to the federal courts, the embodiment of ju- dicial power. In the absence of a Constitu- PROOF: tional issue, however, the presumption is that DD DD JP JP Let w ϵ W /2, w ϵ W /2, and w* ϵ x is not especially low, and the policy deci- xB. Then sion remains in the realm of representative democracy.

RD ϭ JP ϩ ␲ DD ϩ ␲ JP W w M w m w VI. Summary and Avenues for Further ϩ ␲ ͓ ϩ ͑ Ϫ ͒ JP͔ W xw* 1 x w . Research

WDD, WJP and WRD are increasing in x. WJP The paper’s main findings can be summa- is the highest of the three for x ϭ 0, WDD the rized as follows: highest for x ϭ 1, and, for the value x such that WDD ϭ WJP, (1) Accountability has two potential benefits. It allows voters to remove officials whose in- terests appear to be noncongruent with the WRD Ͼ WDD ϭ WJP, electorate, but also gives noncongruent of- ficials some incentive to act as though they since were congruent (through the effect of forward-looking or partial pandering). w* Ͼ wDD ϭ wJP. (2) However, accountability may encourage of- ficials to pander to the electorate and over- Thus, for low values of x or B/L, JP is opti- look minority interests. mal; these are the ranges for which minority (3) Nonaccountability is most desirable when rights are most important, and RD and DD lead (a) the electorate is poorly informed about to the majority-preferred decision too often. For the optimal action, (b) acquiring decision- moderate values of x or B/L, RD is optimal; RD relevant information is costly, and (c) feed- is better than DD because it entails some chance back about the quality of decisions is slow. that the decision will be taken by an m or W Therefore, technical decisions, in particular, 1050 THE AMERICAN ECONOMIC REVIEW SEPTEMBER 2004

may be best allocated to judges or ap- We hope that these extensions and others will pointed bureaucrats. be pursued in future research. (4) The most important decisions should be taken by elected rather than nonaccountable APPENDIX officials (although direct democracy may have the edge over representative democ- We shall suppose that there is a small pro- racy for such decisions). portion ␳ of officials with weak office-holding (5) The discretion of nonaccountable officials motives (so that they will choose actions in the should be more limited than that of account- first period according to their true preferences). able ones. Call these the ideological officials. A proportion (6) Nonaccountability is preferable when the ␲ of these are congruent; the remainder 1 Ϫ ␲ majority’s preferences are very likely to are noncongruent. inflict large negative externalities on the minority. However, representative democ- PROPOSITION A1: When ␦ Ͼ 1 and q ϭ 0, racy is better in this case than direct democ- the unique pure-strategy perfect Bayesian equi- racy, and, for moderate probabilities of librium of RD in the limit when ␳ 3 0 is a pure negative externalities, may constitute a pandering equilibrium in which the official al- desirable compromise between the two ways chooses a in period 1 and is reelected if extremes. and only if she chooses a. The same conclusion holds for mixed-strategy equilibria if we impose This paper is only a first step in the analysis the Markov requirement of footnote 21. of how constitutional design affects public choices. Many other issues of interest come to PROOF: mind. First, extending the model to more than Throughout assume that ␳ Ͼ 0. Suppose, two periods would lead to a richer set of contrary to the proposition, there is a pure- feasible institutions. For example, in a four- strategy equilibrium in which all types of non- period model, the policy of giving an official ideological officials choose b in period 1. Then an initial tenure of two periods and then mak- the choice of a will lead the electorate to believe ing her mandate renewable by vote in each of that the chooser is a congruent ideological of- the last two periods can be shown to make ficial with probability ␲p/[␲p ϩ (1 Ϫ ␲)(1 Ϫ some sense. Second, the analysis could be p)] Ͼ ␲ and so will reelect her. By contrast, the extended to an international context; we could choice of b will lead the electorate to believe get at the idea that elected officials may have that the official is noncongruent with probabil- a harder time establishing credibility interna- ity exceeding 1 Ϫ ␲, and so will not reelect. tionally (e.g., in arms talks or in negotiating Thus, nonideological type (C, a) cannot choose with the IMF) because of their incentive to b in period 1 after all. pander domestically. That is, pandering to Suppose next that there exists a pure-strategy multiple audiences may be difficult. Third, we equilibrium in which some nonideological types could study alternative nomination processes choose a and others choose b. Then, for ␳ suf- for judges and agency commissioners rather ficiently small, the probability of an official’s than maintain our current assumption that being congruent conditional on her choosing a they are simply selected at random. Fourth, in the first period will either be strictly greater we could enrich the model to allow for the than ␲ or strictly less than ␲. In the former case, possibility that campaign contributions affect an official will be reelected if and only if she politicians’ choices. Fifth, the model could be chose a, and in the latter if and only if she chose extended to allow elected officials to “pass b. So, in either case, all nonideological officials the buck” by calling for a public referendum. will have the incentive to behave the same way Finally, some of our analysis might be applied (in order to get reelected), a contradiction. Thus to the media. People often read newspapers the proposition is established for pure-strategy or watch television networks that confirm equilibria. their prejudices; in other words, the me- Next, allow for mixed-strategy equilibria but dia pander in much the same way that politi- impose the Markov requirement. Suppose that cians do. at least one type of nonideological official VOL. 94 NO. 4 MASKIN AND TIROLE: ACCOUNTABILITY IN GOVERNMENT 1051 chooses a with positive probability in equilib- strictly greater than ␲, and so the electorate will rium. Because ␦ Ͼ 1 note that the (C, a) and (N, again reelect the official. b) types have a stronger preference for a over b Suppose instead that a has been revealed to than do the (C, b) and (N, a) types. Hence, in be nonoptimal. If, in equilibrium, no nonideo- equilibrium, the former group chooses a with at logical type chooses a with positive probability least as high probability as the latter group. when a is nonoptimal, the electorate will infer Furthermore, from the Markov assumption, the from the feedback that a noncongruent ideolog- (C, a) and (N, b) types must play a with the ical official has chosen a, and so will not reelect. same probability. Similarly the (C, b) and (N, a) Similarly, if, in equilibrium, some nonideologi- officials must play a with the same probability. cal official chooses a with positive probability Thus, since the proportion of (C, a)s to (N, b)s when a is nonoptimal, then the probability that is ␲p/(1 Ϫ ␲)(1 Ϫ p)(Ͼ␲/(1 Ϫ ␲)), the elec- (N, b) chooses a must be at least as big as the torate will attach a probability greater than ␲ to probability that (C, b) does so [since (N, b)’s the official’s being congruent if a is chosen and preference for a is stronger than that of (C, b)]. so will reelect her. [This is so even if the (C, b) Thus, including the possibility of a noncongru- and (N, a) types choose a with probability 1, ent ideological official, the probability that an because in this case, the ideological officials official is noncongruent conditional on a having will tip the balance in favor of reelection]. Sym- been chosen and revealed nonoptimal is strictly metrically, an official will fail to be reelected if greater than 1 Ϫ ␲, and so the electorate will she chooses b. not reelect the official, establishing the claim for Because officials are reelected if and only if a. The argument for b is entirely symmetric. they choose a, nonideological officials will opt Now consider a nonideological type (N, b) for a, since ␦ Ͼ 1. official. If she chooses b, her payoff, from the above claim, is at least ␦q. If instead she PROPOSITION A2: When q␦ Ͼ (1 ϩ ␦)/ 2, the chooses a, her payoff is at most 1 ϩ ␦(1 Ϫ q). unique limit of perfect Bayesian equilibria as ␳ But, by hypothesis, the former exceeds the lat- 3 0 is an FLP equilibrium in which an official ter. Hence, in equilibrium, type (N, b) cannot is reelected if and only if she has chosen (i) the choose b, and the same for (C, b) [since (C, b)’s optimal action when there is feedback or (ii) preference for b is even stronger]. Similarly, action a if there is no feedback. types (N, a) and (C, a) choose a in equilibrium, establishing that any equilibrium must be FLP. PROOF: Suppose that a has been chosen in the first We claim first that, in equilibrium with ␳ Ͼ 0, period. In equilibrium, nonideological types (C, if the electorate obtains feedback about an offi- a) and (N, a) and ideological types (C, a) and cial’s first-period choice, the official will be re- (N, b) choose a. Hence, if there has been no elected if and only if the decision was optimal. feedback, the probability that the official is con- To see this, suppose that a is the chosen gruent conditional on the choice of a is strictly decision and that the electorate has learned that greater than ␲, and so she will be reelected. it is optimal. If, in equilibrium, no type of Similarly, she will not be reelected if she has nonideological official chooses a with positive chosen b and there is no feedback. probability when a is optimal, the electorate will infer from the feedback that a congruent ϩ ␦ ideological official has chosen a, and so will 1 PROPOSITION A3: When Ͼ q␦ Ͼ 1, reelect. Similarly, if in equilibrium some type of 2 nonideological official chooses a with positive one limit of PBEs as ␳ 3 0 is an FLP equilib- probability, when a is optimal, then the proba- rium in which, if there is no feedback, the elec- bility that (C, a) chooses a must be at least as torate randomizes over reelection. The only big as the probability that (N, a) does so [since other possible limit [if ␦(1 Ϫ 2q) Ն 1] is a full (C, a)’s preference for a is stronger than that of pandering equilibrium. (N, a)]. Thus, if it incorporates the possibility of a congruent ideological official, the probability PROOF: that an official is congruent conditional on a Fix ␳ Ͼ 0. From the proof of Proposition A2, having been chosen and revealed optimal is in any equilibrium, if the electorate obtains 1052 THE AMERICAN ECONOMIC REVIEW SEPTEMBER 2004 feedback about an official’s first-period choice, not reelecting, either when the official has then the official will be reelected if and only if chosen a, or when she has chosen b (or in the decision was optimal. Suppose that there both cases). This implies that the probability exists an equilibrium in which nonideological of an official’s being congruent conditional type (N, a) chooses b with positive probability. on a having been chosen is ␲. Now, the Then, (N, b) and (C, b) both choose b with nonideological types other than (N, b)who probability 1 [since for ␦q Ͼ 1 their preference choose a in equilibrium are (C, a) and (N, a) for b is even stronger than that of (N, a)]. and the ideological types are (C, a) and (N, Hence, without feedback [and in view of the b). Hence, the probability that (N, b) chooses ideological types (N, a) and (C, b) who choose a must be only big enough to offset the effect b], the probability that an official is noncongru- of the ideological officials. Thus, as ␳ 3 0, ent conditional on b having been chosen is equilibrium converges to one in which (N, b) strictly greater than 1 Ϫ ␲, and so an official [and hence (C, b)] plays b with probability who chooses b will not be reelected. This means 1—i.e., an FLP equilibrium—and the elector- that (N, a)’s payoff from choosing b is 1. By ate randomizes over reelection when there is contrast, if she chooses a, she will be reelected no signal, as the proposition claims. (whether or not there is feedback), and so her The only remaining equilibrium possibility is payoff will be ␦, which, by hypothesis, is that nonideological type (N, b) chooses a with greater, a contradiction. We conclude that (N, a) probability 1. Now, if (C, b) does so too, then must choose a with probability 1 in equilibrium, we are done, because this will be a pure pan- which implies that the same is true of (C, a). dering equilibrium. Hence, assume that (C, b) Suppose that there exists an equilibrium in chooses b with positive probability. If this prob- which nonideological type (N, b) chooses b with ability is high enough to outweigh the effect of probability 1. In that case, the probability with- the ideological type (N, a), who chooses b, then, out feedback that an official is noncongruent without feedback, an official who chooses b will conditional on her having chosen b is strictly be reelected, implying that (N, b)’s payoff from greater than 1 Ϫ ␲ (thanks to the fact that, in the b is ␦, whereas that from a is only 1, a contra- absence of feedback, the probability that an diction of the fact that (N, b) chooses a.We ideological official is noncongruent conditional conclude that the probability that (C, b) chooses on her having chosen b is strictly greater than b must be sufficiently small and converge to 0 1 Ϫ ␲), and so (N, b)’s payoff from b is ␦q.By as ␳ 3 0. Thus, in the limit, we obtain a full contrast, if she chooses a, she will be reelected pandering equilibrium, as claimed. if there is no feedback, and so her payoff will be 1 ϩ ␦(1 Ϫ q), which, by hypothesis, is greater, PROPOSITION A4: When ␦ Ͼ 1 and 0 Ͻ a contradiction. We conclude that, in any equi- q␦ Ͻ 1, one limit of PBEs as ␳ 3 0 isaPP librium, (N, b) must choose a with positive equilibrium (which is also a Markov equilib- probability. rium). The only other possible limit [if ␦(1 Ϫ Consider an equilibrium in which nonideologi- 2q) Ն 1] is a full pandering equilibrium. cal type (N, b) randomizes between a and b. Then PROOF: (A1) ␦q ϩ ␦͑1 Ϫ q͒␤ ϭ 1 ϩ ␦͑1 Ϫ q͒␣, Fix ␳ Ͼ 0. Again, it can be shown that, in any equilibrium, an official is reelected when there where the left- and right-hand sides of equation is feedback if and only if her first-period deci- (A1) correspond to the payoffs from b and a sion was optimal. Suppose that there exists an respectively, and ␤ and ␣ are the probabilities of equilibrium in which nonideological type (N, a) reelection in the absence of feedback when, re- chooses b with probability 1. Then the proba- spectively, b and a have been chosen. From equa- bility that an official is noncongruent condi- tion (A1) and hypothesis, we obtain tional on her having chosen b is greater than 1 Ϫ ␲, and so, without feedback, an official 1 Ͼ ␣ Ϫ ␤ Ͼ 0. choosing b will not be reelected. Thus, (N, a)’s payoff from b is 1, whereas her payoff from a is That is, when there is no feedback, the elec- ␦, a contradiction, since the latter is bigger. We torate must randomize between reelecting and conclude that nonideological type (N, a) must VOL. 94 NO. 4 MASKIN AND TIROLE: ACCOUNTABILITY IN GOVERNMENT 1053 choose a with positive probability in equilib- Bohn, Henning and Inman, Robert. “Balanced rium. If she also chooses b with positive prob- Budget Rules and Public Deficits: Evidence ability, then, since 0 Ͻ q␦ Ͻ 1 implies that ⌬ from the U.S. States.” National Bureau of (C, a) Ͼ⌬(N, b) Ͼ⌬(N, a) Ͼ⌬(C, b), type Economic Research (Cambridge, MA). (C, b) will choose b and types (C, a) and (N, b) Working Paper No. 5533, 1996. will choose a: the equilibrium is PP. Notice that Butler, David and Ranney, Austin. Referendums because all four types have different prefer- around the world: The growing use of direct ences, this is also a Markov equilibrium. democracy. Washington, DC: American En- Assume, therefore, that (N, a) chooses a with terprise Institute, 1994. probability 1. Now if nonideological type (C, b) Caillaud, Bernard and Tirole, Jean. “Parties as chooses b with high enough probability to out- Political Intermediaries.” Quarterly Journal weigh the effect of the ideological types (C, b) of Economics, November 2002, 117(4), pp. and (N, a), who choose b, then, without feed- 1453Ð89. back, an official who chooses b will be re- Canes-Wrone, Brandice; Herron, Michael and elected, implying that (N, a)’s payoff from b is Shotts, Kenneth. “Leadership and Pandering: 1 ϩ (1 Ϫ q)␦, whereas that from a is ␦q. But, by A Theory of Executive Policymaking.” hypothesis, the former is bigger than the latter, American Journal of Political Science, 2001, a contradiction. We conclude that nonideologi- 45, pp. 532Ð50. cal type (C, b) can choose b with probability at Carrillo, Juan and Mariotti, Thomas. “Electoral most on the order of ␳. But as ␳ 3 0, (C, b)’s Competition and Politicians Turnover.” Eu- strategy converges to one in which a is chosen ropean Economic Review, January 2001, with probability 1, implying that the limiting 45(1), pp. 1Ð25. equilibrium is full pandering. Dahl, Robert. A preface to democratic theory. Chicago: University of Chicago Press, 1956. Dewatripont, Mathias and Maskin, Eric S. REFERENCES “Credit and Efficiency in Centralized and Decentralized Economies.” Review of Eco- Avery, Christopher and Meyer, Margaret. “De- nomic Studies, October 1995, 62(4), pp. 541Ð signing Hiring and Promotion Procedures 55. When Evaluators are Biased.” Mimeo, KSG Dewatripont, Mathias and Tirole, Jean. “Advo- Harvard and Oxford, 2000. cates.” Journal of Political Economy, Febru- Barro, Robert. “The Control of Politicians: An ary 1999, 107(1), pp. 1Ð39. Economic Model.” Public Choice, Spring Faure-Grimaud, Antoine and Gromb, Denis. 1973, 14, pp. 19Ð42. “Public Trading and Private Incentives.” Beard, Charles. An economic interpretation of Working paper, FMG, 2000. the Constitution of the United States (1913). Ferejohn, John. “Incumbent Performance and New York: MacMillan (1952 edition). Electoral Control.” Public Choice, 1986, Besley, Timothy and Case, Ann. “Political Insti- 50(1Ð3), pp. 5Ð25. tutions and Policy Choices: Evidence from Fudenberg, Drew and Tirole, Jean. “A Theory of the United States.” Journal of Economic Lit- Income and Dividend Smoothing.” Journal erature, March 2003, 41(1), pp. 7Ð73. of Political Economy, February 1995, 103(1), Besley, Timothy and Coate, Stephen. “Non- pp. 75Ð93. Majoritarian Policy Outcomes and the Role Hanssen, Andrew. “The Effect of Judicial Insti- of Citizens’ Initiatives.” Mimeo, LSE, 2000. tutions on Uncertainty and the Rate of Liti- . “Elected versus Appointed Regulators: gation: The Election versus Appointment of Theory and Evidence.” Journal of the Euro- State Judges.” Journal of Legal Studies, Jan- pean Economic Association, September uary 1999, 28(1), pp. 205Ð32. 2003, 1(5), pp. 1176Ð206. . “Independent Courts and Administra- Besley, Timothy and Payne, Abigail. “Judicial tive Agencies: An Empirical Analysis of the Accountability and Economic Policy Out- States.” Journal of Law, Economics, and Or- comes: Evidence from Employment Dis- ganization, October 2000, 16(2), pp. 534Ð71. crimination Charges.” Mimeo, LSS and Mc- Laffont, Jean-Jacques. Incentives and political Master University, 2003. economy. Oxford: Oxford University Press, 1054 THE AMERICAN ECONOMIC REVIEW SEPTEMBER 2004

2000. nomics, November 1997, 112(4), pp. 1163Ð Laffont, Jean-Jacques and Martimort, David. 202. “Separation of Regulators against Collusive Persson, Torsten and Tabellini, Guido. Political Behavior.” RAND Journal of Economics, economics: Explaining economic policy. Summer 1999, 30(2), pp. 232Ð62. Cambridge, MA: MIT Press, 2000. Levy, Gilat. “Legal Precedent and Appellate Re- Rogoff, Kenneth. “The Optimal Degree of Com- view with Careerist Judges.” Mimeo, London mitment to an Intermediate Monetary Tar- School of Economics, 2000. get.” Quarterly Journal of Economics, Madison, James. “Federalist 10,” in Alexander November 1985, 100(4), pp. 1169Ð89. Hamilton, James Madison, and John Jay, Schumpeter, Joseph. (1942) Capitalism, social- eds., The Federalist Papers [1787]. New ism and democracy, 3rd Ed. New York: York: Penguin, 1981. Harper Torchbooks, 1962. Manin, Bernard. The principles of representa- Sie´ye`s, Abbe´. “Dire de l’Abbe« Sie«ye`s sur la tive government. Cambridge: Cambridge Question du Ve«to Royal.” Versailles: Bau- University Press, 1997. doin, Imprimeur de l’Assemble« e Nationale, Matsusaka, John. “Economics of Direct Legis- 1789. lation.” Quarterly Journal of Economics, Stein, Jeremy. “Efficient Capital Markets, In- May 1992, 107(2), pp. 541Ð71. efficient Firms: A Model of Myopic Cor- McCubbins, Matthew D.; Noll, Roger G. and porate Behavior.” Quarterly Journal of Weingast, Barry R. “Administrative Proce- Economics, November 1989, 104(4), pp. dures as Instruments of Political Control.” 655Ð 69. Journal of Law, Economics, and Organiza- von Thadden, Ernst-Ludwig. “Long-Term Con- tion, Fall 1987, 3(2), pp. 243Ð77. tracts, Short-Term Investment and Monitor- Morris, Stephen. “Political Correctness.” Jour- ing.” Review of Economic Studies, October nal of Political Economy, April 2001, 109(2), 1995, 62(4), pp. 557Ð75. pp. 231Ð65. Weingast, Barry R. “The Political Foundations Persson, Torsten; Roland, Gerard and Tabellini, of Democracy and the Rule of Law.” Amer- Guido. “Separation of Powers and Political ican Political Science Review, June 1997, Accountability.” Quarterly Journal of Eco- 91(2), pp. 245Ð63.